Open your bank statement and count the subscriptions that run your business. Booking, invoicing, scheduling, review requests, payroll, cloud storage, plus whatever AI note-taker somebody talked you into at a chamber of commerce mixer. Nine? Twelve?
Now add it up. For a lot of small business owners the total lands between $300 and $500 a month. That's a car payment, except you never finish paying it off. And the rent goes up faster than almost anything else you buy: software subscription prices rose 13.2% over the past year, roughly five times the inflation you see at the grocery store.
Look at that list again, though. Every line on it is a rental, something you pay forever and never own. For about thirty years that was just the cost of doing business, because the alternative, building your own software, was insane.
That last part changed recently, and it's worth walking through the numbers.
The rent never stops
Individually, every tool on the list is cheap. $12 here, $29 there. That's the whole trick, because nobody signs up for $400 of software. You sign up for one $12 tool twelve times, and each one renews on autopilot, forever, whether you opened it this month or not.
Some of them you barely open at all. Zylo analyzed 40 million software licenses for their 2026 index and found that 36% of the licenses companies pay for go completely unused. That's a third of the pile, rented and forgotten.
So the fatigue you feel about software is the correct response to a pile of landlords, not flakiness. For thirty years, the only advice anyone had was to rent less. The idea that some of it should simply be yours wasn't on the table.
The old math was right
In 2015, if the owner of a three-person business told me they were thinking about commissioning custom software, the correct advice was don't. Buy subscriptions instead. Anyone who said otherwise back then was selling you something.
A simple custom app cost $20,000 to $50,000 to build, and something with more moving parts ran $50,000 to $120,000. And that was just to get it built. Software needs feeding, too. Security fixes, updates, small repairs, the annual "the tax rules changed" adjustment. The rule of thumb is 15 to 20% of the build cost, every year. A $40,000 app carried an $8,000-a-year habit.
That math only ever closed one way, by spreading the build cost across thousands of paying users. That's why every software company on earth sells subscriptions. Building software is brutally expensive and copying it is nearly free, so the money only works at scale.
Your five-person business isn't scale, though. "Don't build, buy" was honest arithmetic back then, not the industry pushing you around.
The math is dead
AI got good at writing code, for one. In April, Google's CEO said that three-quarters of all new code written at Google is now generated by AI, with human engineers reviewing it. Take that with a grain of salt. Google sells Gemini and has every incentive to make it sound bigger than it is. Even discounted, the number tells you where the typing went.
The other change is cost. One developer posted about a job he had once quoted at $50,000. The AI bill came to $297. Don't take the number at face value. He went in with years of domain knowledge already in his head, and nobody outside the project can say how good the result is. What I'll promise is the direction. I build small apps for a living and watch these numbers move every month, and quotes that made sense two years ago don't make sense now.
Put those two changes together and the picture looks different for a five-person shop. An app no longer needs thousands of users to justify what it costs to build. An app with five users, your five employees, can make financial sense today. The only test is whether it saves you more than it cost.
There's a category of software that was always needed and never affordable. An app shaped exactly like your business, one where you stop bending your workflow to fit somebody else's tool. "Custom apps that don't scale" sounds like a compromise, but for a five-person shop that's the point.
Try the crude version yourself
Since around 2024, you've been able to sit down with an AI tool, describe what you need in plain English, and get something back that runs. No code required. If a task in your week lives in three spreadsheets and a group text, describe it and ask for the crude version.
If that itch is there, scratch it. The crude version teaches you what you actually need, and every so often it turns out to be enough on its own. Either way, you're only out a weekend.
Go in with open eyes, though. Stack Overflow surveys professional developers every year, and in 2025 66% of them said the most annoying thing about AI coding tools is answers that are "almost right, but not quite". Another 45% said debugging AI-written code takes longer than writing it themselves. The people who do this for a living spend real hours digging out of almost-right, and so will you.
Almost-right is fine for a weekend toy. Less fine when the tool holds your customer list. The crude version has no backups and no real logins, and it lives on one laptop. Nobody is fixing it the Tuesday morning it breaks, right before the weekend rush, and nothing tells you when it quietly stops saving things correctly in March.
None of that is a reason to skip the experiment. Just know which version you have.
Or hand it off
The other fork is that you want the real version. Logins and backups that hold up when it matters, and a human to call when it misbehaves. That's when you hand the whole problem to somebody whose actual job this is.
Same deal I give every consulting client. Tell me what you're trying to fix, and I'll tell you whether software is the right tool for it and roughly what it would cost. If the answer is "keep the spreadsheet," you'll hear that. Some annoyances should just stay annoyances.
What the conversation won't sound like is a $50,000 quote. I wrote before about why those quotes came out ten times too high and what changed. And if you've got an app idea rattling around, start with the reality check before you spend anything.
Nobody offered small custom apps before because the math didn't work. Now it does.
Keep renting, assemble, or build
None of this means you cancel everything tomorrow. When I sort one of these piles with a client, it comes out in three buckets.
Keep renting. Payroll, taxes, bookkeeping, anything where the subscription buys you compliance and someone to blame when the rules change. Owning that category of software would be reckless, like keeping a cow in your backyard because you like lattes. Those subscriptions earn their keep.
Assemble. Some of the pain isn't a missing app. It's five apps that refuse to talk to each other, so you do the talking, retyping the same data between them all month. That's a wiring problem, and wiring is cheap. I wrote a whole post about connecting your tools for free, and for a lot of businesses that fixes most of the annoyance.
Build. Then there's the workflow that is your business. The sequence of steps that makes customers pick you, the one no product models because it's yours. Every off-the-shelf tool makes you bend that workflow to fit. That one is the custom app, and with five users and no scaling it often pays for itself inside a year.
That tired feeling has a simple cause. Everything on the list is a rental, and until recently renting was the only thing on the menu. For the first time in about thirty years, owning your software is priced for a business with five employees.
Start with the ugliest line in your pile, the tool you pay for every month and quietly resent.
If you want a second opinion on which bucket it belongs in, you know where to find me.